Inversion des observations spatiales micro-ondes pour la détermination de la température du sol en présence de neige

Abstract

The soil temperature is an essential parameter for the energy balance of the earth. Many methods have been developed to determine summer surface temperature, but the determination in the presence of snow is an ill-conditioned problem since it requires the differentiation of several temperatures (surface of snow, temperature gradient within the snowpack and temperature at the snow/soil interface). Our project was motivated by the need to improve the estimation of soil temperature, within the first centimeters of soil, under the snowpack.The passive microwave remote sensing could provide this information. We showed the potential of the passive microwave brightness temperature inversion at 10 GHz (derived from AMSR-E, version V5) for the estimation of the soil temperature by using a physical multilayer snow model (SNTHERM) coupled with a snow microwave emission model (HUT).The snow model is driven with measurements from meteorological stations (air temperature, precipitation, air relative humidity, wind speed) and data generated by the NARR meteorological reanalysis.The coupled model is validated with in-situ measurements and the retrieved soil temperatures are compared to those derived from the snow model and NARR.The overall root mean square error in the soil temperature retrieval is 3.29 K, which is lower than the error derived from models without the use of remote sensing. This validation must consider the fact that we are comparing temperatures from a point station to that corresponding to an area of 25 x 25 km on the satellite scale. We also show the possibility of mapping the soil temperature. This original procedure constitutes a very promising tool to characterize the soil under snow (frozen or not), as well as its evolution in locations where measurements are unavailable